Let E(t) be the set of historical information available up until some time t, where t is some date (e.g. 1934). Let p(A|E) be your estimate of the probability an optimally rational Bayesian agent would assign to the event "Self-improving artificial general intelligence is discovered before 2100" given a certain set of historical information.

Consider the function p(t)=p(A|E(t)). Presumably as t approaches 2009, p(t) approaches your own current estimate of p(A).

Describe the function p(t) since about 1900. What events - research discoveries, economic trends, technological developments, sci-fi novel publications, etc, caused the largest changes in p(t)? Is it strictly increasing, or does it fluctuate substantially? Did the publication of any impossibility proofs (e.g. No Free Lunch) cause strong decreases in p(t)? Can you point to any specific research results that increased p(t)? What about the "AI winter" and related setbacks?